The upcoming Oracle Digital Assistant release 23.10 brings us plenty of cool new features, paving the way for even more amazing stuff.
LLM Blocks: You can now add a generative AI capability like summarization, sentiment analysis, and small talk whenever you need it in your dialog flow using the new Invoke Large Language Model (LLM) component. This component handles interactive transactions, such as multi-turn refinements, the back-and-forth exchanges between the user and the LMM that hone the LLM output through rounds of user feedback. You can also use LLM components to collect out for downstream use. The LLM component’s built-in validation provides guardrails against vulnerabilities like prompt-injection attacks that bypass the model’s content moderation guidelines.
https://docs.oracle.com/en/cloud/paas/digital-assistant/whats-new/index.html#GUID-249A3F93-6CC8-4B8F-9B14-EA4F56ECC266
I will soon write a post dedicated to the new LLM component, allowing us to directly integrate with any LLM provider from the dialog flow.
Zoom App channel: There is a new channel for incorporating a digital assistant into Zoom Meetings video conferences through the Oracle Digital Assistant for Zoom app that is available in the Zoom Marketplace
I tried this already and it’s yet another cool feature. We can have ODA “listening” on Zoom meetings and acting as an assistant. I will cover this soon as well.
Automated Data Purge: You can now schedule archive and purge tasks at reqular intervals using the Data Management auto purge preferences. Using these preferences, you can implement your document retention policy’s requirements for automatically deleting data that has fallen outside of the document retention period and create an audit trail that fulfills operational and legal requirements for automated data purging.
New User Tracking in Insights: Starting with this release, User Metrics reporting now tracks new users (the users who have never previously interacted with either a skill or digital assistant).
Bots Node SDK: Support for LLM components and rich form layouts have been added.
MySQL private trust support for SQL Dialogs: You can now register MySQL as a data service using a private CA certificate
SQL Dialogs improvements: In SQL Dialog skills that are based on platform version 23.10, there are the following improvements:
- Abbreviated numbers like “2.5K” are resolved correctly.
- Queries with consecutive values of different type are understood (e.g. “show Chicago salesmen” could be understood as “show employees where the location is Chicago and the job is salesman”.
- Order by multiple attributes (show all employees ordered by salary and hire date).
- Order by multiple conditions (show employees earning more than 1000 and less than 2000).
- Grouped by multiple attributes (show number of employees by job, location, and manager).
Date/Time Enhancements: In user utterances that include a month, the resolution process considers the year value in the context of the utterance. For example, for the phrase “show results for October and December in 2021”, the year for both months should resolve to 2021.
Check the documentation for all the detailed information.